Update PR curve

[glenn-jocher]

This PR update Precision-Recall curve plotting per user requests. PR curves are now colored with mAP shown per class for datasets with up to 20 classes. Larger datasets like COCO plot as grey lines per class, with the same all-class mAP displayed.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhanced precision-recall curve plotting and maintenance improvements in YOLOv5 testing script.

📊 Key Changes

• Updated ap_per_class function to accept save_dir and names parameters instead of fname.
• Added plot_pr_curve function for improved plotting with legend adjustments for a cleaner presentation.
• Modified import comments to reference utils.plots instead of utils.general in various plot functions.
• Streamlined file retrieval in plot_results, using Path objects and eliminating commented-out code.

🎯 Purpose & Impact

• Clearer Visualization: Precision-recall curves can now display individual class APs in the legend, aiding in more detailed analysis.
• Code Organization: Adjusting import comments helps maintain correct module references, reducing confusion.
• File Handling Enhancements: Improved file handling with Path objects for better compatibility and readability in plotting functions.
• User Experience: These changes may lead to a more intuitive understanding of model performance for developers and users analyzing YOLOv5 test results.

[glenn-jocher]

<= 20 classes

> 20 classes

[yasindagasan]

very nice!